Files
ai-video-fullstack/frontend/src/hooks/use-voice-preview.ts
Xin Wang ac3f4dd806 Enhance voice interaction features and introduce voice preview functionality
- Update README to reflect the integration of the DebugVoicePanel with WebSocket support for voice interactions.
- Refactor voice_webrtc.py to improve error handling during WebRTC signaling and include assistant_id in the offer payload.
- Add useVoicePreview hook to manage microphone access and WebRTC connections for real-time voice previews.
- Modify AssistantPage to incorporate new visualizer options and pass assistantId to DebugVoicePanel, enhancing user experience during audio interactions.
- Update API model to include new fields for voice, speed, and language, supporting TTS and ASR configurations.
2026-06-10 10:17:46 +08:00

257 lines
8.0 KiB
TypeScript

"use client";
/**
* 语音预览:把麦克风接到后端 /ws/voice(WebRTC 信令),听到助手实时回应。
*
* 走原生 RTCPeerConnection + 一条 ws 信令通道,与后端 voice_webrtc.py 的约定对齐:
* client → {type:"offer", payload:{pc_id, sdp, type, assistant_id}}
* server → {type:"answer", payload:{pc_id, sdp, type}}
* client → {type:"ice-candidate", payload:{pc_id, candidate:{...}}}
* 音频本身走 WebRTC 媒体流(Opus),不经 ws;后端 TTS 帧从 ontrack 拿到直接播放。
*
* 纯本机(localhost)即可跑:localhost 是 secure context,麦克风可用,ws 用明文。
* 局域网/别的设备要 https+wss,见 deploy/README.md。
*/
import { useCallback, useEffect, useRef, useState } from "react";
import { API_BASE } from "@/lib/api";
export type VoicePreviewStatus = "idle" | "connecting" | "connected" | "failed";
// http→ws、https→wss,自动跟随 API 基址(同源反代时也对)
function wsBaseUrl(): string {
const url = new URL(API_BASE, window.location.origin);
url.protocol = url.protocol === "https:" ? "wss:" : "ws:";
return url.toString().replace(/\/$/, "");
}
function generatePcId(): string {
const bytes = new Uint8Array(16);
crypto.getRandomValues(bytes);
return (
"PC-" +
Array.from(bytes)
.map((b) => b.toString(16).padStart(2, "0"))
.join("")
);
}
type UseVoicePreviewOptions = {
/** 取麦克风失败(权限/无设备)时回调,供 UI 提示。 */
onMicError?: () => void;
};
function errorMessage(error: unknown, fallback: string): string {
if (error instanceof Error && error.message) return error.message;
return fallback;
}
export function useVoicePreview(
assistantId: string | null,
{ onMicError }: UseVoicePreviewOptions = {},
) {
const [status, setStatus] = useState<VoicePreviewStatus>("idle");
const [error, setError] = useState<string | null>(null);
const [localStream, setLocalStream] = useState<MediaStream | null>(null);
const audioRef = useRef<HTMLAudioElement | null>(null);
const pcRef = useRef<RTCPeerConnection | null>(null);
const wsRef = useRef<WebSocket | null>(null);
const localStreamRef = useRef<MediaStream | null>(null);
const startingRef = useRef(false);
const releaseResources = useCallback(() => {
const ws = wsRef.current;
wsRef.current = null;
if (ws) {
ws.onclose = null;
ws.onerror = null;
ws.onmessage = null;
ws.close();
}
const pc = pcRef.current;
pcRef.current = null;
if (pc) {
pc.onconnectionstatechange = null;
pc.onicecandidate = null;
pc.oniceconnectionstatechange = null;
pc.ontrack = null;
pc.close();
}
localStreamRef.current?.getTracks().forEach((track) => track.stop());
localStreamRef.current = null;
if (audioRef.current) audioRef.current.srcObject = null;
startingRef.current = false;
}, []);
const disconnect = useCallback(() => {
releaseResources();
setLocalStream(null);
setError(null);
setStatus("idle");
}, [releaseResources]);
const fail = useCallback(
(message: string) => {
releaseResources();
setLocalStream(null);
setError(message);
setStatus("failed");
},
[releaseResources],
);
const connect = useCallback(async () => {
if (startingRef.current || pcRef.current || wsRef.current) return;
if (!assistantId) {
setError("请先保存助手,再开始语音预览。");
setStatus("failed");
return;
}
startingRef.current = true;
setError(null);
setStatus("connecting");
const pcId = generatePcId();
const ws = new WebSocket(`${wsBaseUrl()}/ws/voice`);
wsRef.current = ws;
ws.onmessage = async (event) => {
try {
const msg = JSON.parse(event.data);
if (msg.type === "answer") {
await pcRef.current?.setRemoteDescription({
type: "answer",
sdp: msg.payload.sdp,
});
} else if (msg.type === "ice-candidate" && msg.payload?.candidate) {
// 后端当前不主动 trickle,留兼容
try {
await pcRef.current?.addIceCandidate(msg.payload.candidate);
} catch {
/* 忽略迟到/重复 candidate */
}
} else if (msg.type === "error") {
fail(msg.payload?.message || "后端无法启动语音会话。");
}
} catch {
/* 非 JSON / 未知消息,忽略 */
}
};
try {
// 1) 等 ws 连上
await new Promise<void>((resolve, reject) => {
ws.onopen = () => resolve();
ws.onerror = (e) => reject(e);
ws.onclose = () => reject(new Error("语音信令连接已关闭。"));
});
// 连上后,信令异常或关闭都结束当前会话并保留失败状态。
ws.onerror = () => {
if (wsRef.current === ws) fail("语音信令连接失败。");
};
ws.onclose = () => {
if (wsRef.current === ws) fail("语音信令连接已断开。");
};
// 2) 建 PeerConnection(纯 STUN,本机/局域网够用)
const pc = new RTCPeerConnection({
iceServers: [{ urls: "stun:stun.l.google.com:19302" }],
});
pcRef.current = pc;
pc.onicecandidate = (e) => {
if (ws.readyState !== WebSocket.OPEN) return;
ws.send(
JSON.stringify({
type: "ice-candidate",
payload: {
pc_id: pcId,
candidate: e.candidate
? {
candidate: e.candidate.candidate,
sdpMid: e.candidate.sdpMid,
sdpMLineIndex: e.candidate.sdpMLineIndex,
}
: null,
},
}),
);
};
pc.ontrack = (e) => {
if (e.track.kind === "audio" && audioRef.current) {
audioRef.current.srcObject =
e.streams[0] ?? new MediaStream([e.track]);
void audioRef.current.play().catch(() => {});
}
};
pc.onconnectionstatechange = () => {
if (pcRef.current !== pc) return;
if (pc.connectionState === "connected") setStatus("connected");
else if (pc.connectionState === "failed")
fail("WebRTC 音频连接失败。");
};
pc.oniceconnectionstatechange = () => {
if (pcRef.current !== pc) return;
const st = pc.iceConnectionState;
if (st === "connected" || st === "completed") setStatus("connected");
else if (st === "failed") fail("WebRTC 音频连接失败。");
else if (st === "disconnected") fail("WebRTC 音频连接已断开。");
};
// 3) 取麦克风 → 加入连接
let stream: MediaStream;
try {
stream = await navigator.mediaDevices.getUserMedia({
audio: {
echoCancellation: true,
noiseSuppression: true,
autoGainControl: true,
},
});
} catch (mediaError) {
onMicError?.();
fail(errorMessage(mediaError, "无法访问麦克风。"));
return;
}
localStreamRef.current = stream;
setLocalStream(stream);
stream.getTracks().forEach((track) => pc.addTrack(track, stream));
// 4) 生成 offer 并发给后端(assistant_id 在 payload 顶层)
const offer = await pc.createOffer();
await pc.setLocalDescription(offer);
const localDescription = pc.localDescription;
if (!localDescription?.sdp) {
throw new Error("浏览器无法创建 WebRTC offer。");
}
ws.send(
JSON.stringify({
type: "offer",
payload: {
pc_id: pcId,
sdp: localDescription.sdp,
type: localDescription.type,
assistant_id: assistantId,
},
}),
);
} catch (connectionError) {
fail(errorMessage(connectionError, "无法连接语音服务。"));
} finally {
startingRef.current = false;
}
}, [assistantId, fail, onMicError]);
// 卸载时收尾
useEffect(() => releaseResources, [releaseResources]);
return { status, error, localStream, connect, disconnect, audioRef };
}